[Previous] [Next] [Up] [Top]

Assignment 1

Multivariate Data Analysis
Psychology 6140


Readings

G&C, plus some one or more of the readings on graphical methods (copies in the Psychology Resource Center, or the 6140 box):
  1. Green & Carroll, Chapter 1 -- overview
  2. W. G. Jacoby, Statistical Graphics for Visualizing Multivariate Data Sage University Paper 120, ISBN: 0761908994 , 1998. [copy in 6140 box, Rm 158]
  3. Wainer & Thissen, "Graphical data analysis", Ann. Rev. Psychology, 1981.
  4. Chambers et al, Graphical methods for data analysis, Chap. 5. [copies in 6140 box, Rm 158]
  5. Friendly, Statistical graphics for multivariate data. SAS SUGI Conference Proceedings, 1991.
  6. Kastellec, J. P. and Leoni, E. L. (2007) Using Graphs Instead of Tables in Political Science, Perspectives on Politics, 5, 755--771.

Supplements

These supplements list sources of further information on statistical graphics and graphical perception.
Data Visualization Gallery
Some examples of the Best and Worst of Statistical Graphics.
Exploratory Data Analysis Short Course
The course notes from my SCS short course
Take this online Graph Design IQ test.
If you can't get a score of 9 or more, there's a problem.
Nutrient Database Explorer
A multidimensional explorer of nutrient data from the USDA. The parallel coordinates displays the nutrient content of foods in the database across 14 dimensions, colored by food group. This highly interactive web presentation by Mike Bostock is a great example of what can be achieved today in visualization of high-dimensional data. You can find many more examples at the main site, bl.ocks.org.

Problems (due: next week)

  1. Literature scan. Examine the literature in your field (a principal journal, over the last year, say), and select several examples using multivariate analysis (2 or more response measures, and/or multiple independent variables). If you can, try to find at least one paper using techniques concerned with relationships between independent and dependent variables (regression, MANOVA, discriminant analysis), and one dealing with dependence among variables or observations (factor analysis, components analysis, cluster analysis, MDS). [You might want to try searching the Electronic Journals in the York University Library.]

    Submit a photocopy of the first page of the paper (with abstract and complete reference); please write the principal analysis techniques and your name on this page. Also, please send me, by email (friendly@yorku.ca), the full reference (authors, title, journal, year, volume, pages) and keywords for the analysis techniques. Use this template, or, type the reference in an email message. Do not send as a Word .doc file. The following format (surprisingly like BibTeX) would be helpful:

     @article{bru:92,
      author = {Edvin Bru and Reidar J. Mykletun and Sven Svebak},
      title = {Neuroticism, Extraversion, Anxiety and Type A Behaviors as
      	Mediators of Neck, Shoulder and Lower Back Pain in Female Hospital Staff},
      year = 1992,
      journal = {Personality and Individual Differences},
        volume = 15,
        number = 5,
        pages = {485-492},
      keywords = {StepReg, Correl},
      note = {Your Name}
      }
    

    Be prepared to discuss:

    1. In each example, what was the substantive problem of interest?
    2. How did the author's use of the technique(s) relate to the content problem?
    3. How were the results of the analyses presented in the paper? What kinds of tables or figures were used to portray the data?
    4. Do other techniques suggest themselves for the problem examined by the author?
  2. Your research. In terms of your own research interest, try to formulate a problem that appears suitable to multivariate analysis.
    1. How would you classify the problem in terms of the system described in class (and in G&C, Ch. 1)?
    2. What other multivariate procedures are suggested by this classification of your problem?
  3. Graphic communication. From the literature you have examined (or something you can find on the web in your area) find one table or graph which you think is a particularly good use of display material to summarize and communicate research findings. Select another which you think is a particularly bad example. Make a photocopy of each, and append a few sentences indicating why you think so for each.
  4. Matrix Algebra Pretest. The purpose of this pretest is purely to give me an idea of the range of background in matrix algebra in the class. Work thru as much of the matrix algebra pretest as you can. Draw a line on the answer sheet at the point where you don't have any idea of the matrix concepts involved.

© Michael Friendly